by Ajay Shah.
On VoxEU, there is a fascinating article titled China and India: Those two big outliers by Jesus Felipe, Utsav Kumar and Arnelyn Abdon.
The interesting fact that they highlight is that both India and China are wise beyond their per capita GDP when it comes to the sophistication and diversification of their exports.
The evidence that they show, on the change in export diversification, is quite striking:
| China | India | |
| 1962 | 105 | 71 |
| 2007 | 265 | 254 |
| Change (times) | 2.52x | 3.58x |
In India's case, in 1962, in the depth of India's autarky, there were 71 commodities exported with `revealed comparative advantage'. By 2007, this number had gone up by 3.58 times. Both China and India are outliers (with excessively high values seen for export diversification) when compared with other countries at the same level of per capita GDP on a PPP basis.
Explaining the unusual export diversification
One element of the explanation of diversification is sheer size. Continental India has a diverse array of locations. Coastal Gujarat is a good location for processing crude oil for export, and Bihar is a good place for growing Litchis for export. By aggregating both places into a single country, we get high levels of export diversification. A casual examination of their graph (Figure 2) makes me think there is some support for this conjecture - positive outliers in the graph are big countries like the US and Germany; negative outliers are small countries like Ireland and Finland.
Explaining the unusual export sophistication
Why does India do sophisticated export, well beyond what one would expect for its level of per capita GDP?
- Sheer size matters. Consider the distribution of a certain
specific kind of knowledge across individuals in the country. Suppose
you set a high cutoff for the minimum knowledge required of that field
in order to assemble a large sized firm. So if you want to build a
large sized firm in that field, you need to recruit 1000 people who
have this specialised knowledge in excess of this cutoff. In a country
of 1.2 billion people, you have more draws from the same
distribution. So even if the lay of the land is quite bad in the sense
that most people have bad knowledge, the sheer size of the country
enables the establishment of firms which require building groups with
high end specialised knowledge.
Consider the distribution of IQ. One in a thousand people have an IQ of above 146. To help fix your intution, it appears that GRE V+Q of 1450 is roughly IQ=146. In India, with a population of 1.2 billion, we have 1.2 million of them. These 1.2 million very smart people in the country can serve as a core around which extremely high quality firms can be built. These effects are accentuated by increasing returns to scale, and the operation of Metcalfe's Law, in the gains from interaction and competition between these people within a country. - There is an odd upper tail in Indian human capital. Looking back
100 years ago, there has been a bizarre upper tail of very highly
skilled people in India. Think Ramanujan: by rights, you would have
never expected that kind of incredible knowledge to be found in a
place like India. But pre-independence India managed to have
incredible geniuses like Ramanujan, C. V. Raman, S. N. Bose and
C. R. Rao -- well before the post-independence push that created the
IITs. Is this merely about size (a lot of draws) or was there actually
a bizarre upper tail?
On this subject, see India shining and Bharat drowning: comparing two Indian states to the worldwide distribution in mathematics achievement by Jishnu Das and Tristan Zajonc. Some fascinating estimates are shown in Producing superstars for the economic Mundial: The team in the tail by Lant Pritchett and Martina Viarengo, who estimate the number of 15 year olds in a country with a OECD PISA score of above 625. The US is estimated to have between 240,000 and 270,000 individuals in this rarefied zone. India has (a) A lot of people, (b) An abysmally poor mode, and (b) A strange upper tail. Putting these together, they estimate India has 100,000 and 190,000 individuals in this rarefied zone - which is incredibly impressive considering that the Indian per capita GDP is one-thirtieth of that seen in the US. This also tells me that we need to scale up the universities in India so that atleast 200,000 individuals each year: it's a shame underutilising these kids.
Aside: PISA > 625 is a much weaker condition than IQ > 146.
Some people bemoan the inequality of human capital that is found in India, i.e. the huge gap between this upper tail and the modal value. But given that we have a low per capita GDP, would we rather have equality where everyone has low skills, or would we rather have an incredible upper tail in the distribution of knowledge, that is able to learn new technology, plug into globalisation, and power the country along?
This is also related to Albert Hirschmann's theme of unbalanced growth: he had argued that growth involves developing an `unbalanced' capability (e.g. India and the software industry led by a small core of high end capabilities), and then harnessing the benefits of the catchup by the rest of system (e.g. telecom reforms, mass scale computer programming education, broad business skills in running globalised firms out of India).
In a recent NBER working paper, Eric A. Hanushek and Ludger Woessmann offer interesting evidence about the tradeoff between `rocket scientists or basic education for all'. They say:
Both the basic-skill and the top-performing dimensions of educational performance appear separately important for growth. From the estimates in column 3, a ten percentage point increase in the share of students reaching basic literacy is associated with 0.3 percentage points higher annual growth, and a ten percentage point increase in the share of top-performing students is associated with 1.3 percentage points higher annual growth
....
the effect of the top-performing share is significantly larger in countries that have more scope to catch up to the initially most productive countries (col. 5). These results appear consistent with a mixture of the basic models of human capital and growth mentioned earlier. The accumulation of skills as a standard production factor, emphasized by augmented neoclassical growth models (e.g., Mankiw, Romer, and Weil (1992)), is probably best captured by the basic-literacy term, which has positive effects that are similar in size across all countries. But, the larger growth effect of high-level skills in countries farther from the technological frontier is most consistent with technological diffusion models (e.g., Nelson and Phelps (1966)). From this perspective, countries need high-skilled human capital for an imitation strategy, and the process of economic convergence is accelerated in countries with larger shares of high-performing students.
Many countries have focused on either basic skills or engineers and scientists. In terms of growth, our estimates suggest that developing basic skills and highly talented people reinforce each other. Moreover, achieving basic literacy for all may well be a precondition for identifying those who can reach “rocket scientist” status. In other words, tournaments among a large pool of students with basic skills may be an efficient way to obtain a large share of high-performers.
On a related note, it is very, very hard to create high end skills when starting from scratch. Witness the difficulties faced by China which had to start from scratch after destroying the elite in the Cultural Revolution. When the economy is ready with demand for a particular set of specialised skills, it may take decades to fill these gaps. As an example, by the late 1980s and early 1990s, it was obvious that there is a giant opportunity for India in software exports and in BPO. But it took 10 years for the education system to re-engineer itself to produce these skills in large quantities, and then make possible large numbers for IT/ITES exports. In similar fashion, the NDA got going on raising expenditure on infrastructure by 2003, but last month, Vikas Bajaj has an article in the New York Times about shortages of civil engineers. It is convenient, in economic development, to have a pre-existing base of high-end skills ahead of time, before the phase of high growth arrives.
Size and economic development
The argument in this blog post has emphasised size. There are many other good things about size, such as economies of scale in the domestic economy, and paying for the fixed costs of global firms in learning about a country in order to do business in it.
If size is such a good thing for economic development, why has it failed so far: as of 2010, why are India and China far behind OECD levels of per capita GDP?
One key story lies in globalisation. Big countries feel they can get away with autarkic policies. They feel self-sufficient and are prone to cut themselves off from the world. Policy makers in small countries don't think they have a choice in trying to create a domestic car industry, but their counterparts in places like Brazil or India or France feel they can experiment with industrial policy. Once this problem is solved -- as seems to be partly the case with India and China where trade liberalisation has arrived though capital account liberalisation has not -- big countries are no longer held back by autarkic policies. In addition, plugging into globalisation, by itself, yields world scale, and thus boosts certain dimensions of size.
Another story, emphasised by Lant Pritchett, lies in the extent to which India is not a single common market, and has thus squandered these potential gains from size. Conversely, as we strip away the legal and tax impediments against intra-India movement of goods, services, capital and labour, and as we bulk up on the infrastructure of transportation and communications, we will obtain returns to size which were not visible in the pre-2000 Indian GDP data.
Finally, on the role of size and sophisticated technological civilisation, see Insufficient data on Charlie's Diary.
I am grateful to Lant Pritchett, Jishnu Das, Pratap Bhanu Mehta, and Josh Felman for comments and improvements on this post.